A replicated state cluster is provided that employs a standby node state assessment during a leadership transition. One method, by a first standby node, comprises monitoring communications from a first cluster leader, wherein the first cluster leader and the first standby node are members of a replicated state cluster; obtaining a designation as a second cluster leader of the replicated state cluster in response to the first cluster leader becoming unavailable; and assessing a state of an additional standby node of the replicated state cluster within a time period after the designation, wherein the additional standby node is available as a candidate leader of the replicated state cluster during the time period when the state of the additional standby node is valid based on one or more validity criteria, wherein the state comprises one or more results of one or more write operations generated by an application.
|
1. A method, comprising:
monitoring, by a first standby node, communications from a first cluster leader executing on a different device than the first standby node, wherein the first cluster leader and the first standby node are members of a replicated state cluster that provides a repository for an application, wherein one or more write operations generated by the application are replicated in the repository across the members of the replicated state cluster;
obtaining, by the first standby node, a designation as a second cluster leader of the replicated state cluster in response to the first cluster leader becoming unavailable; and
assessing, by the first standby node, a state of at least one additional standby node of the replicated state cluster within a time period after the designation, wherein the at least one additional standby node is available as a candidate leader of the replicated state cluster during the time period when the state of the at least one additional standby node is valid based at least in part on one or more validity criteria, wherein the state comprises one or more results of the one or more write operations generated by the application and wherein the assessing comprises comparing a first repository of the at least one additional standby node of the replicated state cluster to a second repository of the second cluster leader of the replicated state cluster;
wherein the method is performed by at least one processing device comprising a processor coupled to a memory.
10. An apparatus comprising:
at least one processing device comprising a processor coupled to a memory;
the at least one processing device being configured to implement the following steps:
monitoring, by a first standby node, communications from a first cluster leader executing on a different device than the first standby node, wherein the first cluster leader and the first standby node are members of a replicated state cluster that provides a repository for an application, wherein one or more write operations generated by the application are replicated in the repository across the members of the replicated state cluster;
obtaining, by the first standby node, a designation as a second cluster leader of the replicated state cluster in response to the first cluster leader becoming unavailable; and
assessing, by the first standby node, a state of at least one additional standby node of the replicated state cluster within a time period after the designation, wherein the at least one additional standby node is available as a candidate leader of the replicated state cluster during the time period when the state of the at least one additional standby node is valid based at least in part on one or more validity criteria, wherein the state comprises one or more results of the one or more write operations generated by the application and wherein the assessing comprises comparing a first repository of the at least one additional standby node of the replicated state cluster to a second repository of the second cluster leader of the replicated state cluster.
16. A non-transitory processor-readable storage medium having stored therein program code of one or more software programs, wherein the program code when executed by at least one processing device causes the at least one processing device to perform the following steps:
monitoring, by a first standby node, communications from a first cluster leader executing on a different device than the first standby node, wherein the first cluster leader and the first standby node are members of a replicated state cluster that provides a repository for an application, wherein one or more write operations generated by the application are replicated in the repository across the members of the replicated state cluster;
obtaining, by the first standby node, a designation as a second cluster leader of the replicated state cluster in response to the first cluster leader becoming unavailable; and
assessing, by the first standby node, a state of at least one additional standby node of the replicated state cluster within a time period after the designation, wherein the at least one additional standby node is available as a candidate leader of the replicated state cluster during the time period when the state of the at least one additional standby node is valid based at least in part on one or more validity criteria, wherein the state comprises one or more results of the one or more write operations generated by the application and wherein the assessing comprises comparing a first repository of the at least one additional standby node of the replicated state cluster to a second repository of the second cluster leader of the replicated state cluster.
2. The method of
3. The method of
4. The method of
5. The method of
6. The method of
7. The method of
8. The method of
9. The method of
11. The apparatus of
12. The apparatus of
13. The apparatus of
14. The apparatus of
15. The apparatus of
17. The non-transitory processor-readable storage medium of
18. The non-transitory processor-readable storage medium of
19. The non-transitory processor-readable storage medium of
20. The non-transitory processor-readable storage medium of
|
The field relates generally to information processing systems, and more particularly, to node clustering in such systems.
In many systems, a cluster comprised of several member nodes manages the system and maintains a repository holding the state of the managed system. For example, a replicated state cluster provides a backup repository for an application using a repository file. As an application writes to the repository file, the write operations are replicated across member nodes of the replicated state cluster. When a new node takes control of the replicated state cluster to become the new cluster leader (e.g., following a leader election), the new cluster leader often declares the other cluster members as unsuitable to assume leadership. This may cause the replicated state cluster to have a single point of failure where only one cluster member is fit to lead the cluster.
A need exists for improved techniques for processing a change in leadership of the replicated state cluster.
In one embodiment, a method comprises monitoring, by a first standby node, communications from a first cluster leader executing on a different device than the first standby node, wherein the first cluster leader and the first standby node are members of a replicated state cluster; obtaining, by the first standby node, a designation as a second cluster leader of the replicated state cluster in response to the first cluster leader becoming unavailable; and assessing, by the first standby node, a state of at least one additional standby node of the replicated state cluster within a time period after the designation, wherein the at least one additional standby node is available as a candidate leader of the replicated state cluster during the time period when the state of the at least one additional standby node is valid based at least in part on one or more validity criteria, wherein the state comprises one or more results of one or more write operations generated by an application.
In some embodiments, the state of the at least one additional standby node is updated during the time period based at least in part on the state of the second cluster leader in response to the state of the at least one additional standby node not being valid based at least in part on the validity criteria. For example, the state of the at least one additional standby node can be updated by synchronizing the state of the at least one additional standby node to the state of the second cluster leader within the time period. Synchronizing the state may comprise evaluating differences between the state of the second cluster leader and the state of the at least one additional standby node and synchronizing only storage blocks that are different between the second cluster leader and the at least one additional standby node. The differences between the state of the second cluster leader and the state of the at least one third device may comprise differences with respect to unacknowledged write operations comprising state information.
Other illustrative embodiments include, without limitation, apparatus, systems, methods and computer program products comprising processor-readable storage media.
Illustrative embodiments of the present disclosure will be described herein with reference to exemplary communication, storage and processing devices. It is to be appreciated, however, that the disclosure is not restricted to use with the particular illustrative configurations shown. One or more embodiments of the disclosure provide methods, apparatus and computer program products for implementing a replicated state cluster with a standby node state assessment during a leadership switchover.
In one or more embodiments, techniques are provided to process a change in leadership of the replicated state cluster by assessing the validity of the other cluster members prior to declaring the other cluster members unsuitable to assume leadership. In addition, when a member is nearly-synchronized with the new cluster leader, techniques are provided to synchronize the nearly-synchronized member with the new cluster leader during the leadership transition, as opposed to declaring the nearly-synchronized member as being out of synchronization with the new cluster leader.
In this manner, a valid cluster can be reestablished, thereby avoiding unnecessary large data updates and potentially losing the cluster leader in case of a failure rendering the entire cluster out-of-synchronization.
As discussed further below, a replicated state cluster typically comprises an odd number of members with a cluster leader node, one or more cluster standby nodes and zero or more tiebreaker nodes (where tiebreaker nodes comprise a voting ability for leader elections but do not participate in the replication of state information). The replicated state cluster replaces a traditional file with cluster redundancy, in a known manner. Thus, the replicated state cluster provides a repository file for an application. The application writes to the repository file which is replicated across the members of the replicated state cluster.
The cluster leader maintains a valid repository and updates the other member nodes so they are synchronized with the cluster leader in case another member node needs to assume leadership of the cluster. When a new leader is elected (e.g., upon startup of a system or the current leader fails or becomes unavailable) using a leader election, the new leader ensures that the other member nodes are synchronized with the new leader and have the same data in their repositories. If a repository of one of the other member nodes is not synchronized with the cluster leader, the outdated member node is typically marked as unsuitable to assume leadership.
For example, in some existing replicated state clusters, when a new member node first becomes the cluster leader, the new cluster leader sets the state of the other member nodes to “degraded” thus marking them as unsuitable to become leaders of the cluster. Thereafter, the new cluster leader initiates a full synchronization of the other member nodes to ensure that the repositories of the other member nodes are identical to the repository of the new cluster leader.
This behavior is inherent in non-journaled leader election algorithms for replicated data. Journal-based protocols, on the other hand, such as the Raft family of protocols, solve consensus in a network of unreliable processors (e.g., processors that may fail). Consensus is the process of agreeing on one result among a group of participants. This problem becomes more difficult when the participants or their communication medium may experience failures. Paxos is another family of consensus protocols that may employ a journal. One or more aspects of the present disclosure are consensus based with a consensus on the member nodes having a valid state and that are eligible to become a leader of the cluster. Among other potential benefits, non-journaled leader election algorithms are typically easier to implement, provide better performance (e.g., with only one write operation, instead of the two write operations for journal-based protocols) and exhibit less overhead (e.g., since no reordering is required) relative to journal-based protocols.
When a new leader takes control, the transactions that were not yet acknowledged to the source application must either be committed or rolled back (recovered transactions). Due to the lack of a journal, these unacknowledged transactions cannot be isolated, so the entire state of the cluster leader must be replicated across all of the member nodes using a full synchronization.
In this full synchronization process, the cluster leader sends the entire repository of the cluster leader to the other member nodes. This process takes a relatively long period of time (e.g., a few seconds) in which the cluster has the current cluster leader as a single point of failure. If the cluster leader experiences a failure (or other unavailability) during the synchronization time, there is no other candidate leader node in the cluster that can assume leadership of the cluster, since the state of the other nodes was set to “degraded” by the prior cluster leader.
As noted above, techniques are provided to synchronize the other cluster members with the new cluster leader during the leadership transition, as opposed to declaring the nearly-synchronized member as being out of synchronization with the new cluster leader. In at least one embodiment, a quick assessment is performed to evaluate the validity of the repositories of the other members during the leadership transition. In this manner, cluster members will not be marked as unsuitable and remain viable as candidate leaders that can potentially assume the role of cluster leader, if needed. For example, if the repositories of other cluster members are identical to the repository of the new cluster leader, the other cluster members will not be marked as degraded. Thus, the number of situations where the cluster will have a single point of failure is reduced by avoiding the degraded state when another cluster member is synchronized with the new leader or can get back to a synchronized state during the leadership transition.
In some embodiments, the repositories of the other cluster members are compared to the repository of the new cluster leader using checksums to assess the synchronization state of the backup repositories and synchronizing only storage blocks that are different between the new cluster leader and the other cluster members. In this manner, the amount of time where the replicated state cluster is exposed to a danger of metadata loss due to having only a single copy of data is reduced and avoids declaring a synchronized replica held by one or more of the other cluster members as being out of synchronization due to a leader transition.
In one or more embodiments discussed herein, when a cluster node becomes the new cluster leader, the new cluster leader sends a start synchronization message to the other cluster members but does not set the state of the other cluster members to a “degraded” state for a predefined grace period for the leadership transition. The predefined grace period may be arbitrarily determined in some embodiments based on the predefined timeouts of the system, on the probability of a failure and/or additional parameters deemed appropriate by a person of ordinary skill in the art. If during the predefined grace period a full synchronization between one or more of the other cluster members and the new cluster leader is reached, the one or more other cluster members will stay in a normal state without moving to degraded (and thus be eligible as standby nodes that may serve as a new cluster leader, if needed).
When the predefined grace period expires, if a synchronization was not completed between the repositories of a given other cluster member and the repository of the new cluster leader, the state of the given other cluster member will be set to “degraded.” In a cluster with two other cluster members, for example, even if only one other cluster member was synchronized, the replicated state cluster will be able to sustain a failure of the current leader at this sensitive stage.
It is noted that, in at least some embodiments, if an additional standby node was “degraded” when the original leader crashed (or otherwise became unavailable), the additional standby node cannot become leader if the new leader crashes during the predefined grace period. Generally, during the predefined grace period, the state that matters is the acknowledged state of the original leader.
The client applications 110 include various types of applications that issue data I/O (input/output) requests to access and store data to and from the data storage devices 152 of the data storage resources 150 of the server nodes 130. For example, the client applications 110 include user applications, server applications, database applications, virtual machines and containers. The client applications 110 can be hosted by, and execute on, various types of computing devices and systems including, but not limited to, desktop computers, laptop computers, workstations, computer servers, enterprise servers, rack servers, smart phones and electronic tablets.
While the communications network 120 is generically depicted in
In some embodiments, the data storage resources 150 comprise direct-attached storage (DAS) resources (internal and/or external storage resources of the server node 130), wherein the storage devices 152 are virtually pooled into shared block storage by the SDS control system. For example, the storage devices 152 include the same type, or a combination of different types of persistent storage devices (e.g., physical block devices) such as hard disk drives (HDDs), solid-state drives (SSDs) (e.g., flash storage devices), PCIe flash cards, or other types and combinations of non-volatile memory. The data storage resources 150 are directly connected to the server node 130 through, e.g., a host bus adapter, and using suitable protocols such as ATA, SATA, eSATA, NVMe, SCSI and SAS. In an exemplary embodiment, the storage devices 152 include both HDD and SSD storage devices. As is known in the art, SSD storage devices provide faster storage performance than HDD devices.
While
In the software-defined storage environment of
In general, a software-defined storage system has certain characteristics which provide the impetus for implementing the replicated state cluster with a standby node state assessment during a leadership transition in accordance with the present disclosure, the functions of which will explained in further detail below.
As used herein, the term “replicated state cluster” comprises an odd number of members with a cluster leader node, one or more cluster standby nodes and zero or more tiebreaker nodes. The replicated state cluster replaces a traditional file with cluster redundancy, in a known manner. Thus, the replicated state cluster provides a repository file for an application. The application writes to the repository file which is replicated across the members of the replicated state cluster.
In the exemplary embodiment described herein, the application that writes operations to the repository file is embodied as a metadata manager that maintains state information for the SDS-based computing system 100 of
In some embodiments, the cluster leader 136 and one or more cluster standby nodes 138-1 through 138-m comprise a replicated state cluster. The various members of the replicated state cluster are in continuous communication (for example, via heartbeat messages) that allow the other members of the replicated state cluster to determine when a cluster member becomes unavailable (potentially triggering a leader election and a leader transition in accordance with the present disclosure).
The SDS control system is a component of the software-defined storage environment shown in
The SDS control system supports the virtualization of storage by separating the control and management software from the hardware architecture. The SDS control system is configured to abstract storage access services from the underlying storage hardware to thereby control and manage I/O requests issued by the client applications 110, as well as support networking and connectivity. As shown in
On the client-side, an SDS data client (SDC) is a lightweight device driver that is deployed in each host application server which hosts one or more of the client applications 110 that require access to the block devices exposed and managed by the SDS control system. The SDC exposes block devices representing the virtual storage volumes that are currently mapped to that host. In particular, the SDC serves as a block driver for a client (server) to access the block storage that is managed by the SDS control system. The SDC provides the operating system or hypervisor (which runs the SDC) with access to the logical block devices (e.g., volumes). The SDCs have knowledge of which SDS control systems (e.g., SDS control system) hold its block data, so multipathing can be accomplished natively through the SDCs. The metadata managers manage SDC to SDS data mapping.
Generally, the metadata manager 134 manages the computing system 100. The metadata manager 134 comprises the metadata required for system operation, such as configuration changes. In one or more embodiments, the metadata manager 134 manages the metadata, the SDCs, SDS, device mappings, volumes, snapshots, system capacity (e.g., device allocations and/or release of capacity), RAID protection, errors and failures, and/or system rebuild tasks (including rebalancing). In addition, user interactions with the computing system 100 are handled by the metadata manager 134 in some embodiments. In a normal input/output (TO) flow, for example, the metadata manager 134 is not part of the data path and user data does not pass through the metadata manager 134. Therefore, the metadata manager 134 is not a performance bottleneck for I/O operations.
In the following discussion, the exemplary metadata manager 134 generates write operations (e.g., the initial state synchronization and ongoing changes) to the cluster leader 136 of the replicated state cluster, and the cluster leader 136 replicates the write operations to the cluster standby nodes 138. In this manner, the replicated state cluster provides a backup storage for the metadata manager 134. In addition, in the following discussion, the exemplary metadata manager 134 receives acknowledgement messages when the write operations are properly received by each member of the replicated state cluster.
In some embodiments, one or more of the cluster standby nodes 138 may be referred to as an “additional standby node,” which can include the cluster leader 136 (who has come back online after being unavailable and can now server as the leader again). In a similar manner, the original cluster leader 136 (prior to a leadership transition) can also be the new leader following the leadership transition.
As noted above, computing system 100 comprises a software-defined storage system that implements the disclosed replicated state cluster techniques with a standby node state assessment during a leadership transition. In one exemplary implementation, the software-defined storage system may be implemented using the Dell EMC ScaleIO® software-defined storage product, commercially available from Dell EMC of Hopkinton, Mass. The Dell EMC ScaleIO™ software-defined storage product is also known as the VxFlex Os® and/or PowerFlex™ software-defined storage product.
While the disclosed replicated state cluster techniques employing a standby node state assessment during a leadership transition are illustrated in one or more embodiments in the context of a software-defined storage system, the disclosed techniques may be employed during a leadership transition for any cluster system, as would be apparent to a person of ordinary skill in the art, based on the present disclosure. For example, clusters are also employed to provide a catalog of the nodes on which particular microservices execute in a microservice based architecture, or to maintain the user permissions of a large system.
In some embodiments of the present disclosure, the members of a replicated state cluster comprise up to three modules, namely, a manager module 250, an actor module 260 and a voter module 270. A key 280 identifies the nature of various communications among the modules 250, 260, 270.
A tiebreaker node 220 runs only the voter module 270 to receive and process votes from other nodes. A standby node 210 runs the voter module 270, as well as the actor module 260 that allows the standby node 210 to receive and process replication messages from the cluster leader. The application node (e.g., the metadata manager 134) that is replicating its state using the replicated state cluster runs the manager module 250 that allows the application to replicate state messages to the nodes in the replicated state cluster 200.
In the example of
It is noted that with existing leader transition techniques, when node B was elected to become the new cluster leader, following the crash of node A, node B would have declared the state of node B as the “truth.” Since the state of node C is not the same as the state of node B in the example of
As shown in
The state of standby node C remains valid, as node B did not declare its own state as the “truth” before crashing (while awaiting the predefined grace period to expire). Thus, node C can become the new cluster leader and the replicated state cluster continues to be available.
During step 830, standby node 210-A is chosen as the new cluster leader following a leader election. At a time 840, standby node 210-B has different data than standby node 210-A. During step 850, standby node 210-A, as the new cluster leader, attempts to synchronize the other standby nodes 210, such as standby node 210-B. At a time 860, a synchronization would normally synchronize the other standby nodes 210 and thereby restore a stable state, allowing the replicated state cluster to remain available. At a time 870, standby node 210-A becomes the new cluster leader and the leadership transition time 880 is over.
It is noted that at time 870, the standby node 210-A becomes the new cluster leader even if synchronization of another standby node 210 has not completed within the predefined grace period.
As shown in
Finally, during step 930, the exemplary cluster leader transition process 900 assesses a state of at least one additional standby node of the replicated state cluster within a time period after the designation. The at least one additional standby node is available as a candidate leader of the replicated state cluster during the time period when the state of the at least one additional standby node is valid based at least in part on one or more validity criteria. In at least some embodiments, the state comprises one or more results of one or more write operations generated by an application, such as the metadata manager 134.
The particular processing operations and other network functionality described in conjunction with the flow diagram of
In at least some embodiments, the cluster leader election is performed in three stages. An initial leader selection stage begins when the leader was elected until the predefined grace period expires or all other member nodes are synchronized. During the initial leader selection stage, no other member of the replicated state cluster may assume leadership and the newly elected cluster leader cannot change the state of the other nodes (e.g., degraded or not).
An inactive leader stage begins after the initial leader selection stage ends and lasts until getting a reply from a majority of the voters in the replicated state cluster that are informed about the new state. During the inactive leader stage, the elected leader may not change a state of the voters, and the elected leader will not perform management operations that may require the repository to change.
If during the initial leader selection stage, the elected leader crashes or a full synchronization was completed, the other member nodes of the replicated state cluster are not degraded, and the repository did not change. Therefore, other standby members of the replicated state cluster, which were valid before the new leader took control, are still valid and can assume leadership. In this manner, the replicated state cluster does not reach a state where there is a single point of failure.
Finally, a leadership stage begins after the inactive leader stage ends and lasts until another new leader is to be elected. During the leadership stage, the elected leader manages the system and may declare a standby node as degraded.
Among other benefits, the disclosed techniques for implementing a replicated state cluster with a standby node state assessment during a leadership transition leverages the uncertainty of distributed systems. For example, the disclosed standby node state assessment during a leadership transition allows a given standby node to become a new cluster leader even though the state of the given standby node is only partially synchronized. In this manner, cluster down time is reduced and availability of the replicated state cluster is increased.
The processors 1002 comprise one or more types of hardware processors that are configured to process program instructions and data to execute a native operating system (OS) and applications that run on the server node 1000. For example, the processors 1002 may comprise one or more CPUs, microprocessors, microcontrollers, application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), and other types of processors, as well as portions or combinations of such processors. The term “processor” as used herein is intended to be broadly construed so as to include any type of processor that performs processing functions based on software, hardware and/or firmware. For example, a “processor” is broadly construed so as to encompass all types of hardware processors including, for example, (i) general purpose processors which comprise “performance cores” (e.g., low latency cores), and (ii) workload-optimized processors, which comprise any possible combination of multiple “throughput cores” and/or multiple hardware-based accelerators. Examples of workload-optimized processors include, for example, graphics processing units (GPUs), digital signal processors (DSPs), system-on-chip (SoC), tensor processing units (TPUs), image processing units (IPUs), deep learning accelerators (DLAs), artificial intelligent (AI) accelerators, and other types of specialized processors or coprocessors that are configured to execute one or more fixed functions.
The storage interface circuitry 1004 enables the processors 1002 to interface and communicate with the system memory 1010, the storage resources 1016, and other local storage and off-infrastructure storage media, using one or more standard communication and/or storage control protocols to read data from or write data to volatile and non-volatile memory/storage devices. Such protocols include, but are not limited to, non-volatile memory express (NVMe), peripheral component interconnect express (PCIe), Parallel ATA (PATA), Serial ATA (SATA), Serial Attached SCSI (SAS), and Fibre Channel. The network interface circuitry 1006 enables the server node 1000 to interface and communicate with a network and other system components. The network interface circuitry 1006 comprises network controllers such as network cards and resources (e.g., network interface controllers (NICs) (e.g., SmartNlCs, RDMA-enabled NICs), Host Bus Adapter (HBA) cards, Host Channel Adapter (HCA) cards, I/O adaptors, and converged Ethernet adaptors) to support communication protocols and interfaces including, but not limited to, PCIe, DMA and RDMA data transfer protocols.
The virtualization resources 1008 can be instantiated to execute one or more services or functions which are hosted by the server node 1000. For example, the virtualization resources 1008 can be configured to implement the various modules and functionalities of the SDS control systems of
A hypervisor is an example of what is more generally referred to as “virtualization infrastructure.” The hypervisor runs on physical infrastructure, e.g., CPUs and/or storage devices, of the server node 1000, and emulates the CPUs, memory, hard disk, network and other hardware resources of the host system, enabling multiple virtual machines to share the resources. The hypervisor can emulate multiple virtual hardware platforms that are isolated from each other, allowing virtual machines to run, e.g., Linux and Windows Server operating systems on the same underlying physical host. The underlying physical infrastructure may comprise one or more commercially available distributed processing platforms which are suitable for the target application.
In another embodiment, the virtualization resources 1008 comprise containers such as Docker containers or other types of Linux containers (LXCs). As is known in the art, in a container-based application framework, each application container comprises a separate application and associated dependencies and other components to provide a complete filesystem, but shares the kernel functions of a host operating system with the other application containers. Each application container executes as an isolated process in user space of a host operating system. In particular, a container system utilizes an underlying operating system that provides the basic services to all containerized applications using virtual-memory support for isolation. One or more containers can be instantiated to execute one or more applications or functions of the server node 1000 as well as execute one or more of the various modules and functionalities of the SDS control systems of
The various software modules of the SDS control systems and the replicated state cluster modules that employ a standby node state assessment during a leadership transition, as shown in
The system memory 1010 comprises various types of memory such as volatile RAM, NVRAM, or other types of memory, in any combination. The volatile memory 1012 may be a dynamic random-access memory (DRAM) (e.g., DRAM DIMM (Dual In-line Memory Module), or other forms of volatile RAM. The non-volatile memory 1014 may comprise one or more of a NAND Flash storage device, an SSD device, or other types of next generation non-volatile memory (NGNVM) devices. The system memory 1010 can be implemented using a hierarchical memory tier structure wherein the volatile system memory 1012 is configured as the highest-level memory tier, and the non-volatile system memory 1014 (and other additional non-volatile memory devices which comprise storage-class memory) is configured as a lower level memory tier which is utilized as a high-speed load/store non-volatile memory device on a processor memory bus (i.e., data is accessed with loads and stores, instead of with I/O reads and writes). The term “memory” or “system memory” as used herein refers to volatile and/or non-volatile memory which is utilized to store application program instructions that are read and processed by the processors 1002 to execute a native operating system and one or more applications or processes hosted by the server node 1000, and to temporarily store data that is utilized and/or generated by the native OS and application programs and processes running on the server node 1000. The storage resources 1016 can include, for example, one or more HDDs and/or SSD storage devices.
It is to be understood that the above-described embodiments of the disclosure are presented for purposes of illustration only. Many variations may be made in the particular arrangements shown. For example, although described in the context of particular system and device configurations, the techniques are applicable to a wide variety of other types of information processing systems, computing systems, data storage systems, processing devices and distributed virtual infrastructure arrangements. In addition, any simplifying assumptions made above in the course of describing the illustrative embodiments should also be viewed as exemplary rather than as requirements or limitations of such embodiments. Numerous other alternative embodiments within the scope of the appended claims will be readily apparent to those skilled in the art.
Luboshitz, Eli, Lieblich, Yuval Peleg
Patent | Priority | Assignee | Title |
11550479, | Oct 27 2021 | Dell Products L.P.; Dell Products L P | Metadata management in storage systems |
11609854, | Oct 28 2021 | Dell Products L P | Utilizing checkpoints for resiliency of metadata in storage systems |
11630773, | Jan 05 2022 | Dell Products L.P.; Dell Products L P | Utilizing a persistent write cache as a redo log |
11650920, | Oct 27 2021 | Dell Products L.P. | Write cache management |
11675789, | Jun 29 2021 | EMC IP HOLDING COMPANY LLC | Tracking utilization of data blocks in a storage system |
11704053, | Apr 06 2022 | Dell Products L.P.; Dell Products L P | Optimization for direct writes to raid stripes |
11789917, | Jan 25 2022 | Dell Products L.P.; Dell Products L P | Data deduplication in a storage system |
11842051, | Jan 25 2022 | Dell Products L.P.; Dell Products L P | Intelligent defragmentation in a storage system |
11868248, | Feb 25 2022 | Dell Products L.P. | Optimization for garbage collection in a storage system |
11921714, | Jul 19 2022 | Dell Products L P; Dell Products L.P. | Managing insert operations of a metadata structure for a storage system |
11960481, | Jun 29 2022 | Dell Products L P; Dell Products L.P. | Managing lookup operations of a metadata structure for a storage system |
11971825, | Jul 14 2022 | Dell Products L.P. | Managing granularity of a metadata structure for a storage system |
Patent | Priority | Assignee | Title |
10445180, | Dec 19 2007 | International Business Machines Corporation | Apparatus and method for managing data storage |
10521311, | Jun 30 2016 | Amazon Technologies, Inc | Prioritized leadership for data replication groups |
5551003, | Dec 11 1992 | International Business Machines Corporation | System for managing log structured array (LSA) of DASDS by managing segment space availability and reclaiming regions of segments using garbage collection procedure |
5764880, | Sep 10 1996 | International Business Machines Corporation | Method and system for rebuilding log-structured arrays |
6052799, | May 15 1998 | International Business Machines Corporation | System and method for recovering a directory for a log structured array |
6941420, | Feb 23 2001 | International Business Machines Corporation | Log-structure array |
7587465, | Apr 22 2002 | Cisco Technology, Inc. | Method and apparatus for configuring nodes as masters or slaves |
9372751, | Sep 06 2012 | International Business Machines Corporation | Free space collection in log structured storage systems |
9514014, | Aug 17 2011 | SCALEIO, LLC | Methods and systems of managing a distributed replica based storage |
20020032835, | |||
20170262519, | |||
20180113640, | |||
20180267893, | |||
20180300075, | |||
20190173739, | |||
20190227845, | |||
WO2019024885, | |||
WO2019024900, |
Date | Maintenance Fee Events |
Mar 26 2020 | BIG: Entity status set to Undiscounted (note the period is included in the code). |
Date | Maintenance Schedule |
Nov 02 2024 | 4 years fee payment window open |
May 02 2025 | 6 months grace period start (w surcharge) |
Nov 02 2025 | patent expiry (for year 4) |
Nov 02 2027 | 2 years to revive unintentionally abandoned end. (for year 4) |
Nov 02 2028 | 8 years fee payment window open |
May 02 2029 | 6 months grace period start (w surcharge) |
Nov 02 2029 | patent expiry (for year 8) |
Nov 02 2031 | 2 years to revive unintentionally abandoned end. (for year 8) |
Nov 02 2032 | 12 years fee payment window open |
May 02 2033 | 6 months grace period start (w surcharge) |
Nov 02 2033 | patent expiry (for year 12) |
Nov 02 2035 | 2 years to revive unintentionally abandoned end. (for year 12) |